matrix-search

matrix-search will be a Full Text Search engine for Matrix, providing improved query searching, performance and ability to be ran locally, with E2E capabilities.

This project uses gb.

Volatility

This project is in rapid-development so things may change entirely day-to-day. This is not something that is usable yet.

Local Daemon

The project is in a working (not user friendly) state.

Both processes require fs access to a common json config file [RO] and [RW] access to a common data directory to write databases to. For these instructions these will both be in the current working directory as config.json and data respectively.

To get started:

1. ensure you have node npm/yarn go>=1.7 gb git installed and working.
2. clone this repository, and enter it
3. create a config file, either based on the sample or using the included utility gen-config (gen-config -h is your friend)
4. build the Go portion by running gb build
5. execute the created executable in the background or in another terminal; e.g ./bin/matrix-search-local --config=config.json --data=data &
6. navigate to the go portion using pushd js_fetcher
7. install dependencies using npm i or yarn
8. navigate back using popd
9. run Node script in the background or in another terminal; e.g node js_fetcher/index.js --config=config.json --data=data &

You will need to run a modified Matrix client which sends /search requests to localhost:8000 riot-web supports these modifications using the matrix-search branches on matrix-js-sdk and matrix-react-sdk and specifying the Search URL using custom_search_url in config.json

To run in a docker container

You'll need to copy config.sample.json to config.json (and replace with appropriate homeserver url, user id, device id and access token or generate using gen-config).

Then you can build and run the docker image like so:

$ docker build -t matrix-search .
$ docker run -v data:/node/js_fetcher/data -t -i -p 8000:8000 matrix-search